| 摘 要: 在商户清分批处理作业中，应用程序的高并发数和单机设备的资源利用率总是有一定上限，数据库本身的处理能力和通讯带宽也对批处理作业有一定的约束。根据清分数据商户原子性特征，设计了一种负载均衡的分片算法，实现清分批处理任务在多应用、多数据库间分布式、高并发协同完成，集群节点还可以线性扩展。通过实验测试，对比使用该算法前后的负载均衡性能，分片算法能够在保证商户原子性的情况下有效均衡清分流水，显著提高清分服务器集群的并发读写性能，从而证明了该算法的有效性。 |
| 关键词: 分布式；高并发；商户原子性；清分 |
|
中图分类号: TP312
文献标识码: A
|
|
| Research on Load Balancing Partition Algorithm of Merchant Sorting based on Multi-database |
|
ZHANG Hui
|
(China UnionPay Merchant Services Co ., Ltd., Shanghai 201203, China)
hz-job@163.com
|
| Abstract: In the merchant sorting batch tasks, high concurrency of applications and resource utilization of single equipment always have a certain upper limit. Processing capacity and communication bandwidth of the database also have certain constraints on batch jobs. According to the atomicity of merchant sorting data, this paper proposes to design a load balancing partition algorithm to realize the distributed and high-concurrency collaborative completion of sorting batch processing tasks among multiple applications and databases. The cluster nodes can also be linearly expanded. Experimental test is made by comparing the load balancing performance before and after using the algorithm. Results verify that the proposed partition algorithm can effectively balance the sorting pipeline under the condition of ensuring the atomicity of merchants, and significantly improve the concurrent read-write performance of the sorting server cluster, which proves its effectiveness. |
| Keywords: distributed; high concurrency; merchant atomicity; sorting |
